http://forums.leaflabs.com/topic.php?id=13458 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:05:19 +0000 http://bbpress.org/?v=1.0.2 q http://forums.leaflabs.com/search.php http://forums.leaflabs.com/topic.php?id=13458&page=2#post-97571 Wed, 09 Oct 2013 05:40:59 +0000 Mike McCauley 97571@http://forums.leaflabs.com/ <p>Same test (whatever it turns out to be: prefer a separate master/slave flag) needs to be used in _i2c_irq_error_handler(). </p> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-97563 Wed, 09 Oct 2013 05:38:59 +0000 Mike McCauley 97563@http://forums.leaflabs.com/ <p>Actually, the test: (dev-&gt;state != I2C_STATE_BUSY) is not a strong enough test to detect slave mode. There are race conditions where an interrupt can occur when acting as a master, but when state is not I2C_STATE_BUSY (eg DONE or ERROR).</p> <p>Really need to have a separate flag to indicate whether it is in slave or master mode. </p> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-36150 Fri, 27 Sep 2013 21:49:11 +0000 Mike McCauley 36150@http://forums.leaflabs.com/ <p>Im also affected by this bug in git master, slowing the port of an open source project to Maple.<br /> Can it please be either fixed soon, or the offending update removed? </p> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-30969 Fri, 20 Sep 2013 14:32:23 +0000 mlundinse 30969@http://forums.leaflabs.com/ <p>The following is my local patch to restore i2c master functionality (tested).<br /> It doesnt touch the slave part so it 'should' keep slave functionality intact (untested).</p> <blockquote><p> @@ -348,8 +348,8 @@ void _i2c_irq_handler(i2c_dev *dev) {<br /> * Add Slave support<br /> */</p> <p>- /* Check to see if MSL master slave bit is set */<br /> - if ((sr2 &amp; I2C_SR2_MSL) != I2C_SR2_MSL) { /* 0 = slave mode 1 = master */<br /> + /* Check to see if that a master transfer is not active */<br /> + if (dev-&gt;state != I2C_STATE_BUSY) { /* Master Xfer and no place else sets I2C_STATE_BUSY */</p> <p> /* Check for address match */<br /> if (sr1 &amp; I2C_SR1_ADDR) { </p></blockquote> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-30957 Thu, 19 Sep 2013 21:56:03 +0000 hitcher404 30957@http://forums.leaflabs.com/ <p>Wow, should have looked at the forums before spending hours figuring this out myself.</p> <p>'ginge' could you at least ask someone to revert your pull request? </p> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-28634 Fri, 30 Aug 2013 11:25:52 +0000 mlundinse 28634@http://forums.leaflabs.com/ <p>I think there must be a separate state flag to show if we are making a master or slave transfer. </p> http://forums.leaflabs.com/topic.php?id=13458&page=2#post-28633 Fri, 30 Aug 2013 11:13:22 +0000 ginge 28633@http://forums.leaflabs.com/ <p>confirmed. as the original submitter of the slave code, I will try and get to the bottom of this over as soon as possible. </p> http://forums.leaflabs.com/topic.php?id=13458#post-28631 Fri, 30 Aug 2013 10:45:26 +0000 mlundinse 28631@http://forums.leaflabs.com/ <p>The following small sketch crashes but works if I remove the slave part of the interrupt handler. The borad is connected to a BMP050 barometric pressure sensor.</p> <p><code><br /> #include &lt;stdint.h&gt;</p> <p>#include &lt;libmaple/i2c.h&gt;</p> <p>#define I2CPORT I2C2</p> <p>/************************************************************<br /> *<br /> * I2C convenience functions<br /> *<br /> ************************************************************/</p> <p>i2c_msg msg[2];<br /> uint8_t buffer[20];</p> <p>void i2c_read_buffer(uint8_t device, uint8_t address, uint8_t * buf, uint8_t count)<br /> {<br /> int32_t rc;<br /> /* registar address (W)*/<br /> msg[0].addr = device;<br /> msg[0].flags = 0;<br /> msg[0].data = (uint8_t *)&address;<br /> msg[0].length = 1;<br /> msg[0].xferred = 0;<br /> /* count byte's data request (R)*/<br /> msg[1].addr = device;<br /> msg[1].flags = I2C_MSG_READ;<br /> msg[1].data = buf;<br /> msg[1].length = count;<br /> msg[1].xferred = 0;<br /> rc = i2c_master_xfer(I2CPORT,msg,2,1000);<br /> if (rc&lt;0) {<br /> I2CPORT-&gt;error_flags = 0;<br /> I2CPORT-&gt;state = I2C_STATE_IDLE;<br /> }<br /> }</p> <p>/*******************************************************************</p> <p>Application - setup and main loop</p> <p>*******************************************************************/</p> <p>void setup() {</p> <p> SerialUSB.begin();<br /> i2c_init(I2CPORT);<br /> i2c_master_enable(I2CPORT, I2C_BUS_RESET);</p> <p> int16_t value;<br /> i2c_read_buffer(0x77, 0xAA, buffer, 2);<br /> value = (buffer[0]&lt;&lt;8) + buffer[1];</p> <p>}</p> <p>void loop() {<br /> if (SerialUSB.available()&gt;0) {<br /> SerialUSB.read();<br /> int16_t value;<br /> i2c_read_buffer(0x77, 0xAA, buffer, 2);<br /> value = (buffer[0]&lt;&lt;8) + buffer[1];<br /> SerialUSB.println(value);<br /> }<br /> }<br /> </code> </p> http://forums.leaflabs.com/topic.php?id=13458#post-28629 Fri, 30 Aug 2013 10:20:58 +0000 mlundinse 28629@http://forums.leaflabs.com/ <p>Its a big sketch with lots of other stuff in it but I will try to prepare a minimal example. I think any I2C master read transfer will break, but I havn't had time to test this yet.</p> <p>Its not an OS issue but I am using the latest updated github libmaple code and an Olimex Maple clone, compiling under Fedora 18 with gcc version 4.7.4 20130613 (release) [ARM/embedded-4_7-branch revision 200083].</p> <p>The i2c slave support was merged into the main leaflabs git repository on 26 of july so it is a rather recent issue, and it appeared for me after merging the latest leaflabs patches into my repository a few days ago.</p> <p>I have used a debugger (JTAG) to find the exact line in the code where the crash happens.</p> <p>470:/work/hardware/leaflabs/cores/maple/libmaple/i2c.c **** dev-&gt;i2c_slave_msg-&gt;data[0] = dev-&gt;regs-&gt;DR;</p> <p>This is in the newly added I2C slave handling code in the interrupt handler, and since I only do master transfers, the slave part of the handler should not be reached.<br /> Once the slave section is reached the code will try to write the read data into the slave message structure which is not defined since we are in a master read, Bang-Crash :(</p> <p>Regards<br /> Magnus </p> http://forums.leaflabs.com/topic.php?id=13458#post-28625 Fri, 30 Aug 2013 09:17:22 +0000 crenn 28625@http://forums.leaflabs.com/ <p>Can you post the sketch which exhibited this issue? I haven't seen this occur before. Also what libmaple/Maple IDE version were you using and what OS and board? </p> http://forums.leaflabs.com/topic.php?id=13458#post-28619 Fri, 30 Aug 2013 04:23:11 +0000 mlundinse 28619@http://forums.leaflabs.com/ <p>The latest work on I2C, great work really, breaks the Master I2C mode.<br /> The problem seems to be that after a master receive operation the MSL bit<br /> in I2C_SR2 is reset to 0 when the stop condition is detected, and this happens<br /> before the ISR can handle the received master data.</p> <p>Changing line 352 in i2c.c from:<br /> <code>if ((sr2 &amp; I2C_SR2_MSL) != I2C_SR2_MSL) { /* 0 = slave mode 1 = master */</code></p> <p>to:<br /> <code>if (0) { /*ALWAYS MASTER */</code></p> <p>thus disabling all slave interrupt handling, restores master functionality in my I2C sketch.</p> <p>Regards<br /> Magnus </p>